IETF 118 PCE WG Rough Notes

Thursday, 9 November 2023, Session III, 15:00 - 16:30 CET

• Chairs: Dhruv Dhody, Julien Meuric
• Secretary: Andrew Stone
• Note Taker: Andrew Stone
• Slides
• Video
• Chat-Stream
• Meetecho Recording

Introduction

1.1. Administrivia, Agenda Bashing (Chairs, 5 min) [5/90]

1.2. WG Status (Chairs, 10 min) [15/90]

1.3. State of WG I-Ds, open issues and next steps (Chairs, 10 min)
[25/90]

• Various drafts on schedule with dedicated time slots. Few documents
  nearing WGLC such as draft-ietf-pce-flexible-grid and
  draft-ietf-pce-segment-routing-policy-cp which is key document and
  hoping to work through the process.
• Authors suggested to contact chairs if their document in adoption
  queue needs expedited process (with justification).

Stateful PCE

2.1 Native-IP (Aijun Wang, 10 min) [35/90]
draft-ietf-pce-pcep-extension-native-ip-26

• [Dhruv] - thanks for update, as shepherd thanks for editing from
  comments. Want WG to have eyes on the latest changes (non-editorial)
  such as encoding is changing. Tunnel mode - not really sure if
  described well, and whether we should call it a tunnel. What would
  be the right terminology to use (such as instead of Tunnel Mode?). I
  see this still pending. Taking Status field example, errors and fail
  approach can be much simpler such as error code carried separately.
  Avoid overloading status field. Some ideas I have and wanted to
  share.
• [Aijun] We will try to add some description for Tunnel Mode in the
  document. For Error field, will consider it.

2.2 Stateful Inter-Domain (Julien Meuric/Olivier Dugeon, 5 min)
[40/90]
draft-ietf-pce-stateful-interdomain-04

• [Dhruv] - there was discussion in the past about error handling,
  we parked it because we did not have any of the extensions using it
  and thought this document would be an ideal use case for it - so
  please check that.
• [Olivier] - Yes, will add in and look for in next release as well
  as implementation section as ODL implementation ongoing.
• [Dhruv] - awesome to hear. With regards to PCEP going down, such
  as state sync, any impact with that? Might need more descriptions
  about things going up, down, etc. Distribute label seems clear,
  cleanup of other protocol specs probably to work next.
• [Olivier] - yes, can look at more detail about recover in case of
  failure.

2.3 Inter Stateful PCE State Sync (Cheng Li, 5 min) [45/90]
draft-ietf-pce-state-sync-05

• [No comments]

Segment Routing

3.1 SID Algo (Samuel Sidor, 5 min) [50/90]
draft-ietf-pce-sid-algo

• [Dhruv] - regarding wglc, don't need all code points to do wglc.
  i.e don't worry to need to do wglc -after- allocation.
• [Samuel] - and feel free to contact me directly about
  implementations.

3.2 SR P2MP Policy (Hooman Bidgoli, 10 min) [60/90]
draft-ietf-pce-sr-p2mp-policy-04

• [Adrian] - wondered about objective functions. Original p2mp PCE
  defined two objective functions for shortest path and least cost I
  think, do you see the need for any other objective functions for SR
  P2MP.
• [Hooman] - good question, since it's controller focused, think
  providers will also look at latency too. When it comes to streaming,
  like mobile video streaming of live shows a lot of delay of video
  being transferred. I would see latency coming in.
• [Adrian] - yeah pce can do what it wants for computation, but
  entity asking to create the path, what type of path does it lead to
  ask for.
• [Dhruv] - checked, can use any metric defined. With terminology,
  borrowing instance id from RFC 3209, got confused looking for that
  reference. We need to decide as a WG if we are okay with CP having a
  concept of path instance inside of the candidate path because in
  normal p2p we have a Candidate Path with segment list and list of
  segment. I don't follow the reasoning. We should discuss a bit more.
• [Hooman] - honestly, memory, going back 2 years when we started
  this I think the instance we came up, maybe it's incorrect, thought
  we came up with it as part of the tree instance itself.
• [Dhruv] - we have it everywhere so if this instance is right word
  or not worried about it
• [Hooman] - way we identify the tree is root id, tree id and the
  instance
• [Dhruv] - yep, fine with design, just the term
• [Hooman] - gotcha
• [Dhruv] - when we talk about slicing, need to be careful. In PCEP
  let's talk about flex algo etc. and the realization techniques, we
  have to be clear it's an extension.
• [Cheng li] - I read it but found it complicated. Do we have any
  implementation of this now?
• [Hooman] - up to now implementation have been cli but now PCEP is
  ongoing to try and get an implementation.
• [Cheng li] - recommend add implementation status.
• [Hooman] - we will add the implementation section before WGLC.

Others

4.1 PCEPS update for TLS 1.3 (Sean Turner, 10 min) [70/90]
draft-ietf-pce-pceps-tls13-01

• [Sean] - doc changed quite a bit, should we do another WGLC?
• [Adrian] - decision to do another WGLC will leave to chairs, I
  just wanted to say thank you for doing the security work and coming
  here with this.
• [Julien] - no harm for another WGLC.
• [Sean] - it's (doc) ready to go in data tracker already.

4.2 Fine-grain Transport Network (Liyuan Han, 10 min) [80/90]
draft-han-pce-path-computation-fg-transport-00

• [Julien] - thank you for presentation. As far as I understood
  there's a new dataplane specified in ITU-T, how much of this work
  has been discussed with CCAMP? Whatever we do here, need to do in
  coordination with CCAMP. Any plan with CCAMP? Any schedule or draft?
• [Liyuan] - for this meeting we also requested timeslot with CCAMP
  but no available time. Draft has been sent to that WG.
• [Julien] - encourage you to trigger discussions with CCAMP.
• [Liyuan] - okay will continue on to discuss with CCAMP.

4.3 Precision Availability Metrics (Luis M. Contreras, 10 min)[90/90]

draft-contreras-pce-pam-00

• [Samuel] - about metric type extension, isn't it fixed length?
• [Julien] - yes it is, so it can't be the one as it is as 5440. It
  would need to be new code point
• [Dhruv] - i understand we define a new metric type. I think best
  to have a generic extended metric type like various tlvs. So we
  solve this in one passing go.
• [Dhruv] - another thought, I haven't read PPM draft, but use cases
  from other parts of the world, am I using this with SR Policy or ?
  Where would we see and use this? It provides some insights what
  metric can be useful.
• [Luis] - yep sure will work on adding that
• [Olivier] - goal is to cover a close loop system? Why do you need
  this interval ratio and threshold and why not directly recompute a
  path incase of eelay constraints are not met. For example, original
  request, if path is asking for <20ms delay, if at a certain moment
  parameters change, path exceeds it, decide to recompute. Why do you
  need this ratio and interval etc and not trigger path computation
  immediately
• [Luis] - good question, point is to take into account historical
  behavior. If we keep historical how it was behaving, then we can
  check that. Recompute incase of problems could also be checked.
  Selecting or picking the path based on historical observed. In
  otherwords, two different paths with same metric right now might
  have different historical behavior. I can choose from the more
  adequate path based on the history.